DisposableHeart's Suggested Method
Don’t begin! Folding is highly addictive. But if you do, try out my way of folding: Secondary Structure First of all, I like the loop representation of the protein, because it gives me much more control during the tweaks (If hydrogen bonds would actually be stronger than rubber bands, and could be frozen at will, many times showing the other structures would help.) The sequence below indicates how I use the techniques as I advance to the final state. Strategy Sequence 1. In structural view I examine the structure to find sheets and helices. Sheets should be aligned parallel when possible to let them form Hydrogen bonds. I found helices feel comfortable when their axes are parallel and the winds are little interlaced. (It is a tuning option to snap them along on each other to get them even more comfortable.) Usually I use bands to tie sheets together. The first Wiggle usually pulls them together, when the Hydrogen bonds take over and act more like a band. Hydrogen bonds are strong while present, but unlike bands, they tend to pop with some tension. After this stage the basic structure of the protein is determined. 2. After this phase I use Structure Mode to make everything loops. 3. Global wiggle, local shake, local wiggle repeated as long as possible. I have to maintain the Hydrogen bonds all along. 4. Pull the Backbone to create some Clashes, shake, wiggle, shake, wiggle, while there is improvement. 5. Pull the Sidechains, one by one, more at once, causing clash, or without causing clash, then wiggle. After causing this type of clash shaking is not a good option, because it just releases the tension caused simply by turning the side chain to its initial position. 6. After a while pulling the backbone effectively is not possible. Pulling is either ineffective, or simply cannot pull so little that it wouldn’t cause too many clashes. If there are too many clashes, the structure just explodes after the wiggle. In this case I use bands to deform the shape to create clashes. As soon as there is a clash I delete the bands and use shake and wiggle, local or global in any sequence, repeatedly. This is the direct tuning of the shape of the backbone. 7. Until this stage I have not used any locks yet. (You may have noticed the local shake and local wiggle before. Well, I applied them on the whole. They behave differently than the global actions.) 8. Now the local pull stage comes. I lock certain part of the structure, usually about 6 to 10 units of the backbone, and pull the locked part. Usually I use both global and local wiggle and shake to release the tension I caused with the pull. This is repeated on different sections of the protein while there is score increase. This may take hours. Sometimes I go back to Step #3 and test the structure randomly for Stability. I like when the molecule (at least) returns to the highest score after applying Step #4. Local pull with locks can be applied with pulling side chains around the same area. It happens many times, that I pull the locked part of the backbone, the score drops, but when after this I wiggle the whole structure, nothing happens. In this case I have to make another bigger move that suddenly creates much tension in the structure and makes the original pull effective. The reason for this is that the FoldIt application makes full modeling of the atomic weights and forces. It must, otherwise it could not work. Little forces cause little acceleration, many times below the noise level (if you know what I mean). The tension created by the little pull is not enough to get the structure move, but when a much bigger tension is present, the rearranging structure takes the little forces into account, too. You may have noticed that wiggling structures cannot be pulled directly. They can be pulled with a band though! 9. In this step I make some random locks and do local wiggles, repeatedly. I go back to Step #3 to ease my mind before I go to the next, one of the most tedious methods. 10. Worm wiggle. I begin at one end of the backbone. Lock the 3rd unit, two are free. This lock will be the head of the worm. Make a local wiggle. If there is no score increase, I undo the wiggle, or restore the best. (There is difference. There are some quick transitions at the beginning of the wiggle that are not visible and it seems there is only loss. Restore best can catch these momentary high scores.) If there is improvement, I am happy and continue the walk. I take up 3 units and wiggle and evaluate the result as before. Then lock the first unit of the backbone. This is the tail of the worm. There are two units free. Wiggle and evaluate. Move the head by one unit. There are three units free. Wiggle and evaluate. Move the tail. I repeat until I reach the other end of the protein. I found that wiggling more than 3 units usually just drops the score. I don't know why. Maybe this is puzzle-dependent. 11. After the Worm wiggle I usually go back to Step #3 and quickly tease the molecule again with the already used methods. Sometimes there is significant improvement with the global techniques, and as an obvious consequence, Worm wiggle gives a lot of gain again. 12. An optional step comes: I rebuild both ends of the protein, sometimes a few short sections in the middle. Never more than 5 units. This tool is not really a well-thought one of the FoldIt application. Not only because it is not documented, but there is too much visual effect and useless undo functionality. It should not display much and should remember incremental results only, not any weaker ones. This is a lottery at this version of the application. But, of course, Step #3 releases the pain. 13. The following method is for fine tuning only. I call it micro pull. It is similar to the worm wiggle, but instead of wiggling the small sections, I pull them. A little. In various directions. Usually I move the cursor along a spiral trajectory. In some directions there is a score increase. I play with the direction and the force. Restore the best and move along. This method should not increase the score by more than a pixel (0.05 point) per action. If that happens to you and it happens often, then you are not at the fine tuning stage yet: do something else. General notes: I use the relative score coloring. It shows the location when there is the most score one can gain. I use the “Show All” side chain option almost all the time. I do not use the voids. Not because it is slow, it is not that slow, but I didn’t find it useful. When I pull a side chain into a void, the score drops and the side chain feels better at the original location. It would be good, if the application displayed the TOTAL VOLUME of VOIDS. This would tell how compact the molecule is. (But I guess this is equivalent with the score.) I do not care about the hydrophobic side chains. I could never ever intentionally hide any of them. Somehow increasing the score eventually and inevitably will herd them inside. I save often. It happened that I reached a relatively high score, but my app crashed and I had to start over. I found I am better at some structures than at others. Or the other way around. Like if molecules have personality. Peter Peto alias DisposableHeart Category:Strategy